Early Decrease in Serum Amphiregulin Or Vascular Endothelial Growth Factor Levels Predicts Sorafenib Efficacy in Hepatocellular Carcinoma

ONCOLOGY REPORTS 41: 2041-2050, 2019

Early decrease in serum amphiregulin or vascular endothelial growth factor levels predicts sorafenib efficacy in hepatocellular carcinoma

CORINNE GODIN1,2*, SANDRA BODEAU2,3*, ZUZANA SAIDAK1,4, CHRISTOPHE LOUANDRE2, CATHERINE FRANÇOIS5, JEAN‑CLAUDE BARBARE6, ROMAIN CORIAT7,8, ANTOINE GALMICHE1,2 and Chloé SAUZAY1,2

1Equipe CHIMERE, EA 7516, Université de Picardie Jules Verne; 2Laboratoire de Biochimie; 3Laboratoire de Pharmacologie; 4Service d'Oncobiologie Moléculaire, Centre de Biologie Humaine, CHU Sud; 5EA4294, Université de Picardie Jules Verne; 6Délégation à la Recherche Clinique et à l'Innovation, CHU Sud, 80054 Amiens; 7Gastroenterology and Digestive Oncology Unit, Cochin University Hospital, APHP; 8Unité Inserm U1016, Université Paris Descartes, Sorbonne Paris Cité, 75006 Paris, France

Received June 29, 2018; Accepted November 15, 2018

DOI: 10.3892/or.2018.6922

Abstract. Sorafenib is the standard of care for the treatment at the transcriptional and post‑transcriptional levels. All HCC of advanced hepatocellular carcinoma (HCC). However, patients in our cohort had detectable concentrations of AREG identifying secreted biomarkers that predict sorafenib efficacy and VEGF both at baseline and after sorafenib treatment. The in all HCC patients remains challenging. It was recently decreased serum levels of AREG and VEGF after 15 days of reported that sorafenib interferes with protein homeostasis sorafenib treatment were significantly associated with better and inhibits global translation in tumour cells. A likely overall and progression‑free survival. The results of the consequence of this inhibition would be the interruption multivariate analysis demonstrated that a decrease in AREG of autocrine loops. The aim of the present study was to was an independent prognostic indicator of overall survival investigate the effect of sorafenib on two growth factors (hazard ratio, 0.208; 95% confidence interval, 0.173‑0.673; implicated in autocrine loops and HCC tumour invasion: P=0.0003). These results suggest that sorafenib inhibits auto amphiregulin (AREG) and vascular endothelial growth crine loops and that early decrease in serum AREG or VEGF factor (VEGF). ELISA, quantitative polymerase chain levels predicts sorafenib efficacy in HCC patients. reaction analysis, western blotting and a cytokine array were performed on HCC cell lines and the prognostic role of these Introduction two biomarkers in HCC patients was evaluated. Serum AREG and VEGF levels were assayed by ELISA in 55 patients with Liver cancer is the fifth most prevalent cancer and the second advanced HCC treated with sorafenib. It was observed that most frequent cause of cancer‑related mortality worldwide (1). sorafenib decreased AREG, VEGF and cytokine expression Hepatocellular carcinoma (HCC) is the most frequent type of primary liver tumour. The majority of patients are diagnosed at an advanced stage, which explains the poor prognosis of this cancer. Sorafenib is the standard of care for the treat- Correspondence to: Dr Chloé Sauzay, Laboratoire de Biochimie, ment of advanced HCC. Sorafenib is a multikinase inhibitor Centre de Biologie Humaine, Avenue René Laennec, Centre de directed against the RAF kinases and several receptor tyrosine Biologie Humaine, CHU Sud, 80054 Amiens, France kinases, such as vascular endothelial growth factor receptor E‑mail: sauzay.chloe@chu‑amiens.fr (VEGFR) (2). The efficacy of sorafenib was demonstrated

* a decade ago in two comparative placebo‑controlled trials Contributed equally that showed a benefit in terms of overall survival (OS) (3,4). Recently, lenvatinib demonstrated a survival advantage Abbreviations: AREG, amphiregulin; EGFR, epidermal growth factor receptor; OS, overall survival; PFS, progression‑free survival, that was non‑inferior to that of sorafenib in an open‑label, UPR, unfolded protein response; VEGF, vascular endothelial growth multicentre, non‑inferiority, randomised trial (5). However, factor sorafenib currently remains the only first‑line treatment approved by the Food and Drug Administration for patients Key words: amphiregulin, vascular endothelial growth factor, with advanced HCC. sorafenib, hepatocellular carcinoma, prognostic biomarker, There is currently a lack of biomarkers able to predict the proteostasis efficacy of sorafenib in HCC patients. α‑Fetoprotein (AFP) is the biomarker most widely used by clinicians for the follow‑up of HCC patients treated by sorafenib. However, 2042 GODIN et al: DECREASE OF AREG OR VEGF PREDICTS SORAFENIB EFFICACY IN HCC

AFP is secreted in only ~50% of HCC patients due to high were taken into consideration. This cut‑off was based on the intratumoural variability (6,7). Among these patients, only study of Personeni et al (8). 37% exhibit a significant decrease in their AFP levels following sorafenib treatment (8). Recently, the latest recommendations Determination of AREG and VEGF concentrations. AREG of the European Association for the Study of the Liver or VEGF concentrations in cell supernatant or serum have underlined the suboptimal performance of AFP as samples were determined using a sandwich‑based ELISA kit a serological test in the surveillance of HCC patients (1). (R&D Systems, Inc., Minneapolis, MN, USA) according to the Identifying a biomarker secreted in all patients with HCC that manufacturer's instructions. Only >5% decreases in the serum is directly regulated by sorafenib remains a challenge. In a level of AREG or VEGF were taken into consideration This recent study exploring the impact of sorafenib on proteostasis, cut‑off was selected as it corresponded to the smallest varia- we reported that sorafenib inhibits global protein biosynthesis tion observed in our cohort. and initiation of translation (9). Proteostasis may be defined as the process regulating the production, folding, trafficking Reverse transcription‑quantitative polymerase chain reac‑ and degradation of proteins within the cell, in order to tion (RT‑qPCR) analysis. Total RNA was extracted using maintain its homeostasis. In HCC cells exposed to sorafenib, RNeasy Mini kit (ref 74104; Qiagen, Courtaboeuf, France) this inhibition of translation is expected to largely prevent the and reverse‑transcribed to cDNA using High Capacity cDNA synthesis of several biologically active peptides and proteins. Reverse Transcription kit (ref 4368814; Applied Biosystems; Although this putative effect has not been formally examined, Thermo Fisher Scientific, Inc., Waltham, MA, USA) a likely consequence of protein synthesis inhibition would be according to the manufacturer's instructions. Amplification interruption of autocrine loops (10). was performed with the TaqMan Universal PCR master Mix Proliferation of HCC cells is dependent on two well‑docu (ref 4304437; Applied Biosystems; Thermo Fisher Scientific, mented autocrine loops composed of VEGF/VEGFR and Inc.) on an ABI 7900HT Sequence Detection System amphiregulin (AREG)/epidermal growth factor receptor (Applied Biosystems; Thermo Fisher Scientific, Inc.) using (EGFR) ligands and receptors. VEGFR is a direct target of primers and probe sets for AREG (ref Hs00950669_m1, cat. sorafenib and is known to be an important factor in HCC no. 433182; Applied Biosystems; Thermo Fisher Scientific, growth (11). VEGF is a potent angiogenic factor and is Inc.) and glyceraldehyde phosphate dehydrogenase (GAPDH; upregulated in several human tumours (12). This autocrine loop ref Hs99999905_m1, cat. no. 4333764; Applied Biosystems; plays a crucial role in tumour invasion and aggressiveness (12). Thermo Fisher Scientific, Inc.). The cycling conditions were Activation of EGFR during hepatocarcinogenesis is not initial denaturation at 95˚C for 10 min, followed by 45 cycles dependent on activating mutations, but rather on autocrine of denaturation at 95˚C for 15 sec and annealing at 60˚C loops involving natural ligands of EGFR, such as AREG (13). for 1 min. Data were normalized to the endogenous control AREG is upregulated in HCC and acts as a pro‑oncogenic GAPDH to obtain ΔCq. The fold change in genes of interest factor (14). This autocrine loop was recently identified as a key relative to untreated samples was determined using the determinant of resistance to sorafenib (15,16). The aim of the 2‑ΔΔCq method (17). present study was to evaluate the effect of sorafenib on AREG and VEGF production in two HCC cell lines. In order to study Western blotting. For each experimental condition, complete a potential link between AREG and VEGF production and the cell extracts were prepared in RIPA buffer. Following protein efficacy of sorafenib, the prognostic impact of each of these concentration determination with a BCA kit (Thermo Fisher factors was investigated in HCC patients. Scientific, Inc.), a total of 50 µg protein was precipitated with methanol and chloroform. The samples were then denatured Materials and methods in Laemmli sample buffer, loaded on 10% SDS‑PAGE, and transferred to nitrocellulose membranes using standard proce- Cell culture and reagents. The human HCC cell lines used in dures. The membranes were saturated for 1 h in 5% milk in this study (Huh7 and Hep3B, passages between 6 and 8) were TTBS [Tween 0.05%, NaCl 200 mM, Tris‑HCl (pH 8.0)], then obtained from Dr Wychowski (Institut de Biologie de Lille, rinsed and incubated overnight with each primary antibody at Lille, France) and were authenticated by profiling short tandem a 1:1,000 dilution (AREG, mouse, Santa Cruz Biotechnology, repeats at 16 loci (LGC Standards, Strasbourg, France). The Inc., Dallas, TX, USA; ref 74501; and β‑actin, mouse, cells were cultured in Dulbecco's modified Eagle's medium Sigma‑Aldrich; Merck KGaA; ref A544R). Later, secondary supplemented with 10% fetal calf serum (Jacques Boy antibodies coupled with horseradish peroxidase (HRP) Institute S.A., Reims, France), 2 mM glutamine, penicillin and were incubated for 1 h at a 1:5,000 dilution (anti‑mouse IgG streptomycin. Brefeldin A was purchased from Sigma‑Aldrich; HRP‑conjugated, sheep, GE Healthcare, Chicago, IL, USA; Merck KGaA (Darmstadt, Germany). Sorafenib was purchased ref NA931V). The electrochemiluminescence reaction was from Selleck Chemicals (Houston, TX, USA) and stored as used for detection. 10 mM stock in dimethyl sulfoxide at ‑20˚C. Cytokine array. The Proteome Profiler ‘Human Cytokine Array Determination of AFP concentrations. AFP concentra- Panel A’ kit was purchased from R&D Systems (ref ARY005) tions were measured in serum samples using the Vista and used according to the manufacturer's instructions. Briefly, Dimension 500 analyser (Siemens AG, Munich, Germany) this kit utilizes capture antibodies spotted onto a nitrocellulose and the corresponding kit recommended for routine clinical membrane to allow high‑throughput multi‑analyte profiling practice. Only decreases of >20% in the serum level of AFP of 36 cytokines, chemokines and acute phase proteins in a ONCOLOGY REPORTS 41: 2041-2050, 2019 2043

Figure 1. Sorafenib inhibits AREG, VEGF and cytokines in vitro. (A) Huh7 and Hep3B cells were treated with sorafenib (10 µM) for 18 h. The concentrations of secreted AREG were determined by ELISA in the cell supernatant. (B) Hep3B cells were treated with sorafenib (10 µM) or brefeldin A (BFA; 5 µg/ml) for 18 h. The concentrations of secreted and intracellular AREG were determined by ELISA and the ratio was calculated after normalization with respect to control conditions. (C) qPCR analysis of the mRNA encoding AREG in Hep3B cells exposed to sorafenib (10 µM) for 18 h. (D) Protein extracts obtained from Hep3B cells exposed to 10 µM sorafenib at indicated times were analysed by immunoblotting for their content of AREG or actin. The AREG/actin ratio is given for each time point, taking the control conditions as reference. (E) Huh7 and Hep3B cells were treated with sorafenib (10 µM) for 18 h. The concentrations of secreted VEGF were determined by ELISA in the cell supernatant. (F) Human cytokine array performed on Hep3B cells treated with sorafenib (10 µM) for 18 h. Experiments were performed in triplicate. *P<0.05 compared with control (Student's t‑test). AREG, amphiregulin; VEGF, vascular endothelial growth factor; qPCR, quantitative polymerase chain reaction. single sample. Cell culture supernatants were diluted and the Declaration of Helsinki regarding ethical principles for mixed with a cocktail of biotinylated detection antibodies. medical research involving human subjects. The ‘Comité de The sample/antibody mixture was then incubated with the Protection des Personnes d'Ile de France’ Institutional Review array. Any cytokine/antibody complex present was bound by Board approved the study protocol (NuAT140) and all patients its cognate immobilized capture antibody on the membrane. provided their written informed consent. No samples were Streptavidin‑HRP and chemiluminescent detection reagents obtained from patients who were minors or physically or were added, and a signal was produced in proportion to the mentally unable to understand and provide their consent to the amount of cytokine bound. Chemiluminescence was detected use of their serum samples. in the same manner as a western blot. Statistical analysis. Student's t‑test was used for statistical Serum samples. Frozen serum samples from a cohort of analyses of the experiments performed on cells. Wilcoxon's 55 patients with advanced HCC receiving sorafenib were used test, Fisher's exact test, Chi‑squared test, receiver operating in this study. Patients were recruited from Cochin Hospital characteristic (ROC) curve analysis and Kaplan‑Meier (Paris, France). Of the 55 patients, 50 had cancer associated analysis were used as indicated for analyses performed on with cirrhosis. The patients were treated with sorafenib at the the patient cohort. Univariate and multivariate analyses were validated dose of 400 mg twice daily until evidence of disease performed using a Cox proportional hazard regression model. progression. In 12 of the 55 patients, the dose was reduced to P<0.05 was considered to indicate a statistically significant 50% of the planned dose due to severe drug‑related adverse difference. All analyses were performed with R, v.3.4.2 soft- events (grade 3/4 hand‑foot skin reaction, n=5; grade 3 fatigue, ware (https://www.r‑project.org). n=4; and grade 3 hypertension, n=3). Tumour evaluation was performed every 3 months during treatment according to the Results modified Response Evaluation Criteria In Solid Tumours (18). Progression‑free survival (PFS) was measured from the date Sorafenib inhibits the production of AREG, VEGF and of diagnosis to the date of evidence of disease progression or inflammatory cytokines by HCC cells. In order to analyse death. OS was measured from the date of diagnosis to the date the impact of sorafenib on AREG and VEGF, two cell lines of death or the last follow‑up. Serum samples were collected with different sensitivities to sorafenib were exposed to a from each patient 7 days before and 14 days after initiating clinically relevant concentration of the drug (10 µM) for 18 h. sorafenib therapy. The concentration of 10 µM was selected as it was close to the median concentrations measured in the serum of HCC Ethics approval and consent to participate. The present study patients by Abou‑Alfa et al (19). The Huh7 cell line has been was conducted in compliance with the French legislation and previously shown to be sensitive to sorafenib, while Hep3B 2044 GODIN et al: DECREASE OF AREG OR VEGF PREDICTS SORAFENIB EFFICACY IN HCC

Table I. Summary of the clinical characteristics of the patients. Prognostic role of AFP, AREG and VEGF in HCC patients prior to sorafenib treatment. The clinical characteristics of Characteristics HCC patients (n=55) the patients are summarized in Table I and are also detailed in a previous study (21). The median OS assessed by the Age (years, median) 61 Kaplan‑Meier method was 23.6 months for all 55 patients Sex (male/female) 40/15 enrolled in this study. AFP concentrations were determined Child‑Pugh class in the serum samples of 38 patients among the cohort of 0 12 55 patients, and the concentrations in 19 patients were unde- tectable (<10 ng/ml). By contrast, detectable concentrations A 41 of AREG and VEGF were measured in the 55 patients of B 2 our cohort. In order to evaluate the prognostic value of AFP, Main aetiology of cirrhosis AREG and VEGF prior to treatment, patients were sepa- Alcohol 4 rated according to their baseline AFP, AREG and VEGF NASH 11 concentrations as below or above the median. Kaplan‑Meier Virus 21 analysis of OS demonstrated that baseline AFP, AREG or Mixed 4 VEGF concentrations prior to sorafenib therapy did not Unknown 10 constitute a discriminant prognostic biomarker (Fig. 2A‑C; P=0.10, P=0.98 and P=0.91, respectively). A marked inter‑ HCC, hepatocellular carcinoma; NASH, non‑alcoholic steatohepatitis. and intra‑individual heterogeneity was observed between AFP, AREG and VEGF levels before and after sorafenib treatment (data not shown). Among the 19 patients with detectable concentrations of AFP, pretreatment AFP cells were found to be resistant to sorafenib in terms of concentrations were not significantly different from the clonogenic growth (15). Recently, the Huh7 and Hep3B cell concentrations obtained after 2 weeks of sorafenib therapy lines were exposed to increasing concentrations of sorafenib (2,581±1,053 vs. 1,935±740 ng/ml; P=0.62). Similarly, among (0‑20 µM) for 18 h, and a decrease in cell viability of ~10% the 55 patients of the cohort, pretreatment AREG and VEGF was measured using the Trypan blue exclusion assay (9). concentrations were not significantly different from the This modest cytotoxic effect is essentially accounted for by concentrations obtained after 2 weeks of sorafenib therapy ferroptosis, a form of necrotic cell death (20). (60.6±103 vs. 59.6±105 pg/ml, P=0.91 for AREG; and A significant decrease in AREG concentrations in the 516.6±51 vs. 495.0±43 pg/ml, P=0.75 for VEGF). cell supernatant was observed following sorafenib exposure in both cell lines (Fig. 1A). In order to examine the relative Prognostic role of a decrease in serum AFP levels. Among contribution of defective production vs. secretion of AREG, the 19 patients with detectable concentrations of AFP, we the concentration of this protein was measured in cell observed a significant decrease in AFP (>20%) after sorafenib supernatants and cell lysates following exposure of Hep3B treatment in only 3 patients. OS and PFS were evaluated in cells to sorafenib. The secretion blocker brefeldin A was patients with a decrease in AFP levels vs. patients with no used as positive control. The ratio of secreted to intracellular decrease after 2 weeks of sorafenib therapy. The median AREG was calculated under control conditions or after survival was 31.7 months in patients with a decrease in AFP exposure to sorafenib or brefeldin A. Following sorafenib (n=3) vs. 21.3 months in the group with no decrease in AFP exposure, this ratio was calculated to be 0.60, suggesting that levels (n=16). The difference between these two groups was the inhibitory effect of sorafenib on AREG resulted from not significant (Fig. 2D, P=0.20). The PFS in patients with a both inhibition of AREG production (60%) and inhibition of decrease in AFP was not significantly different from that in AREG secretion (40%) (Fig. 1B). In order to confirm the effect patients with no decrease in AFP levels (Fig 2E, median 27 of sorafenib on AREG production, the effect of sorafenib vs. 18 months; P=0.22). on AREG mRNA levels was examined and AREG protein levels were determined by western blot analysis using Hep3B A decrease in serum AREG or VEGF levels predicts sorafenib cells. An inhibition of 65% in AREG mRNA expression efficacy. In order to evaluate the prognostic value of intra‑indi was observed following exposure to sorafenib (Fig. 1C). A vidual variations of AREG and VEGF, patients were classified decrease of ~50% in AREG protein levels was measured according to their VEGF or AREG variations following by western blot analysis (Fig. 1D). A significant decrease sorafenib therapy. No significant difference was observed of VEGF levels in the cell supernatant was also observed between these groups (presented in Table II for AREG and following sorafenib exposure in both cell lines (Fig. 1E). Table III for VEGF). OS and PFS were evaluated in patients Finally, in order to explore the effect of sorafenib on several with a decrease in one of these biomarkers (>5%) vs. patients cytokines, a cytokine array was performed. Inhibition of with no decrease, after 2 weeks of sorafenib therapy. The three cytokines, namely CXCL1, IL8 and SERPIN E1, was median survival was 38.9 months in patients with a decrease observed (Fig. 1F). Sorafenib exerted an inhibitory effect in AREG (n=26) vs. 15.8 months in the group with no decrease on AREG, VEGF and cytokine expression in vitro at the in AREG (n=29). The difference between these two groups transcriptional and post‑transcriptional level, suggesting that was statistically significant (Fig. 2F; P=0.0009). A statistically sorafenib induced a broad inhibition of bioactive protein significant difference in PFS was also observed between these production. two groups (Fig. 2G, median 46.9 vs. 21.1 months; P=0.0014). ONCOLOGY REPORTS 41: 2041-2050, 2019 2045

Figure 2. Prognostic role of AFP, AREG and VEGF in HCC patients treated by sorafenib. (A-C) Kaplan‑Meier analysis of OS of patients divided into two groups according to their baseline (A) AFP, (B) AREG of (C) VEGF levels (below or above the median). (D) Kaplan‑Meier analysis of the OS of patients divided into two groups according to their variations in AFP before and after sorafenib therapy (presence or absence of a ≥20% decrease of AFP after treatment). (E) Kaplan‑Meier analysis of the PFS of patients divided into two groups according to their variations in AFP before and after sorafenib therapy (presence or absence of a ≥20% decrease of AFP after treatment). (F) Kaplan‑Meier analysis of the OS of patients divided into two groups according to their variations of AREG before and after sorafenib therapy (presence or absence of a ≥5% decrease of AREG after treatment). (G) Kaplan‑Meier analysis of the PFS of patients divided into two groups according to their variations of AREG before and after sorafenib therapy (presence or absence of a ≥5% decrease of AREG after treatment). (H) Kaplan‑Meier analysis of the OS of patients divided into two groups according to their variations of VEGF before and after sorafenib therapy (presence or absence of a ≥5% decrease of VEGF after treatment). (I) Kaplan‑Meier analysis of PFS of patients divided into two groups according to their variations of VEGF before and after sorafenib therapy (presence or absence of a ≥5% decrease of VEGF after treatment). AFP, α‑fetoprotein; AREG, amphiregulin; VEGF, vascular endothelial growth factor; HCC, hepatocellular carcinoma; PFS, progression‑free survival; OS, overall survival.

Interestingly, the group of patients with no decrease in AREG and cirrhosis aetiology, was evaluated by univariate and was composed of 1 patient with a stable level of AREG and multivariate analysis for OS (Table IV). Univariate analysis 28 patients with an increase in AREG levels of at least 20% indicated that a decrease in AREG levels was significantly after 15 days of sorafenib treatment. Therefore, as shown in associated with OS (HR=0.341; P=0.002). Further multivariate Fig. 2F and G, an increase in AREG levels was correlated with analysis demonstrated that a decrease in AREG levels could a poor prognosis. be considered as an independent prognostic biomarker for OS For VEGF, the median survival was 27.2 months in patients (HR=0.208; P=0.00034; 95% CI: 0.173‑0.673). with a decrease in VEGF (n=23) vs. 22.2 months in the group In order to confirm these results, the clinical usefulness with no decrease in VEGF (n=32). A significant difference of a decrease in AREG or VEGF for prediction of OS status in OS was observed between these two groups (Fig. 2H; at 1, 3 and 5 years was evaluated by time‑dependent ROC P=0.0311). A significant difference was also observed in curve analyses. A decrease in AREG was predictive for OS at PFS (Fig. 2I, median 20.9 vs. 16.9 months; P=0.028). The 5 years, with an area under the curve (AUC) of 0.77. The AUC prognostic value of each clinicopathological factor, including was 0.68 at 1 year. For VEGF, AUCs of 0.67 and 0.57 were AREG decrease, VEGF decrease, age, sex, Child‑Pugh score obtained for OS at 5 years and at 1 year, respectively (Fig. 3). 2046 GODIN et al: DECREASE OF AREG OR VEGF PREDICTS SORAFENIB EFFICACY IN HCC

Table II. Summary of the clinical characteristics of the patients categorized according to their variation in serum levels of AREG after 14 days of sorafenib treatment.

Decrease of AREG Decrease of AREG Characteristics (n=29) (n=26) P‑value

Age (years, median) 61 61 0.97a Sex (male/female) 20/9 20/6 0.72b Child‑Pugh class 0.35c 0 6 6 A 23 18 B 0 2 Main aetiology of cirrhosis 0.16c Alcohol 0 4 NASH 6 5 Virus 11 10 Mixed 3 1 Unknown 6 4 aWilcoxon's test; bFisher's test; cChi‑squared test; AREG, amphiregulin; NASH, non‑alcoholic steatohepatitis.

Table III. Summary of the clinical characteristics of patients categorized according to their variation in serum levels of VEGF after 14 days of sorafenib treatment.

No decrease of VEGF Decrease of VEGF P‑value Characteristics (n=32) (n=23)

Age (years, median) 61 62 0.76a Sex (male/female) 21/11 19/4 0.22b Child‑Pugh class 0.77c 0 6 6 A 25 16 B 1 1 Main aetiology of cirrhosis 0.73c Alcohol 2 2 NASH 6 5 Virus 13 8 Mixed 1 3 Unknown 7 3 aStudent's t‑test; bFisher's test; cChi‑squared test; VEGF, vascular endothelial growth factor; NASH, non‑alcoholic steatohepatitis.

The regulation of AREG and VEGF by sorafenib is potentially biomarker alone. The variations of these biomarkers were mediated via a common pathway. To investigate the hypothesis compared in each patient. Interestingly, the two markers varied of a common pathway regulating AREG and VEGF, the in the same direction (decrease or increase) in 69% of the prognostic value of variations of both AREG and VEGF were patients (n=38) (Fig. 4C). A Chi‑squared test revealed that the evaluated. The median OS assessed by the Kaplan‑Meier variations of these two biomarkers were dependent (Fig. 4D; method in patients with a decrease in both AREG and VEGF P=0.007), suggesting that AREG and VEGF are regulated by was 40.5 vs. 21.9 months in the group with an increase or sorafenib via a common mechanism. discordant results for these two markers (Fig. 4A; P=0.0085). A significant difference was also observed in PFS (median Discussion 23.6 vs. 17 months, P=0.017; Fig. 4B). These results suggest that the combination of these two biomarkers was not more The present study demonstrated that sorafenib decreased discriminant in terms of prognosis compared with each AREG, VEGF and cytokine levels in HCC cells at the ONCOLOGY REPORTS 41: 2041-2050, 2019 2047

Table IV. Univariate and multivariate analyses of overall survival in HCC patients.

Univariate analysis Multivariate analysis ------Variables HR 95% CI P‑value HR 95% CI P‑value

Age (years) 1.013 0.989‑1.038 0.291 1.006 0.978‑1.035 0.683 Sex (male) 1.323 0.690‑2.538 0.399 1.342 0.642‑2.807 0.434 Child‑Pugh A 1.162 0.556‑2.427 0.690 0.423 0.061‑2.938 0.384 Child‑Pugh B 7.968 1.565‑40.560 0.012 6.681 0.494‑90.398 0.153 Cirrhosis 1.170 0.453‑3.017 0.746 1.295 0.109‑15.351 0.838 Aetiology of cirrhosis Alcohol 3.287 0.834‑12.948 0.089 8.405 1.606‑43.997 0.012 Virus 1.318 0.481‑3.610 0.592 1.583 0.478‑5.239 0.452 NASH 0.700 0.231‑2.124 0.529 1.008 0.255‑3.990 0.991 Mixed 1.913 0.507‑7.223 0.339 NA NA NA AREG decrease 0.341 0.173‑0.673 0.002 0.208 0.088‑0.491 0.0003 VEGF decrease 0.583 0.312‑1.092 0.092 0.720 0.360‑1.438 0.352

Bold print indicates statistical significance. HCC, hepatocellular carcinoma; HR, hazard ratio estimated from Cox proportional hazard regression model; CI, confidence interval of the estimated HR; NASH, non‑alcoholic steatohepatitis; AREG, amphiregulin; VEGF, vascular endothelial growth factor; NA, not available.

Figure 3. Prognostic value of a decrease in AREG or VEGF evaluated by time‑dependent ROC curve analyses. (A-C) ROC curve analyses for prediction of OS status at 1, 3 and 5 years for AREG decrease. (D-F) ROC curve analyses for prediction of OS status at 1, 3 and 5 years for VEGF decrease. AREG, amphiregulin; VEGF, vascular endothelial growth factor; ROC, receiver operating characteristics; OS, overall survival. transcriptional and post‑transcriptional levels. These results inhibition in a clinical setting. In the present study, a decrease suggest that sorafenib induced an inhibition of the production in serum AREG and VEGF levels was identified as a potential of biologically active proteins, in accordance with our recent biomarker of sorafenib efficacy in HCC patients. study highlighting a global inhibitory effect of sorafenib on Sorafenib is the standard of care for the medical treatment protein biosynthesis (9). This previous study was performed of advanced‑stage HCC, but there is a lack of biomarkers to in vitro and, therefore, did not explore the consequences of this evaluate its effectiveness. AFP is the serum biomarker most 2048 GODIN et al: DECREASE OF AREG OR VEGF PREDICTS SORAFENIB EFFICACY IN HCC

Figure 4. Variations of AREG and VEGF levels are correlated in HCC patient survival. (A) Kaplan‑Meier analysis of the OS of patients with a decrease of both AREG and VEGF levels compared with the other patients. (B) Kaplan‑Meier analysis of the PFS of patients with a decrease of both AREG and VEGF compared with the other patients. (C and D) Comparison of the concordance of the direction of the variations of AREG and VEGF concentrations in patients. AREG, amphiregulin; VEGF, vascular endothelial growth factor; HCC, hepatocellular carcinoma; OS, overall survival; PFS, progression‑free survival.

commonly used by clinicians to evaluate sorafenib efficacy. It These biomarkers may help clinicians identify patients for was previously reported that a decrease of 20% in the serum whom sorafenib is not expected to be beneficial and to consider levels of AFP measured during the first 8 weeks of sorafenib therapeutic alternatives at an earlier stage. A prospective study treatment was associated with better OS, while imaging was not comparing the prognostic role of AFP, AREG and VEGF on a predictive in this context (8). However, a significant decrease in larger number of patients is required to confirm these results. serum AFP levels following sorafenib treatment is observed in Previous studies have investigated the prognostic role of only ~18% of the patients (8). In our cohort, we observed such AREG and VEGF in HCC patients. Our conclusions regarding a decrease in only 3 of 38 patients. Due to the small number VEGF confirm and extend those of another previous study of patients, a decrease in serum levels of AFP was not found by Tsuchiya et al. In complete agreement with our results, to be associated with a statistically significant improvement in these authors observed that a decrease in VEGF after OS or PFS in our cohort (Fig. 2D and E, P=0.20 and P=0.22, 8 weeks of sorafenib therapy was associated with better OS respectively). These results confirm that the clinical value of on Kaplan‑Meier analysis (22). Three studies have already AFP is limited to a small number of patients. explored AREG as a potential prognostic biomarker in HCC In the present study, a decrease in serum AREG or VEGF patients, two of which evaluated AREG at baseline and one levels after only 15 days of sorafenib treatment was shown to after sorafenib treatment. In agreement with the two studies predict sorafenib efficacy. In particular, Kaplan‑Meier analyses evaluating AREG at baseline, we did not observe any correla- revealed that a decrease in serum AREG or VEGF levels after tion between pretreatment AREG levels and prognosis (23,24). 2 weeks of sorafenib treatment was associated with a significantly In the present study, although large fluctuations were observed better OS and PFS. A difference was observed for VEGF and in AREG concentrations in patients, pretreatment AREG AREG in the multivariate analysis. A decrease in AREG, but concentrations were not statistically significantly different not in VEGF, was a significant prognostic biomarker for OS. from the concentrations measured after 2 weeks of sorafenib This difference was also observed with the ROC curve analyses. therapy (Fig. 2A). This may be explained by the marked A decrease in AREG was predictive of OS at 5 years with an inter‑ and intra‑individual heterogeneity in AREG levels AUC of 0.77, whereas the AUC was 0.67 for VEGF. Based on before and after sorafenib treatment. The study by Blivet‑Van these results, we propose that a decrease in both serum VEGF Eggelpöel et al is the only study to have investigated variations and AREG levels may be considered as a prognostic biomarker, in serum AREG levels before and after sorafenib therapy. In but it is likely that the decrease in serum AREG levels is a better contrast with the present study, these authors demonstrated biomarker in this context. Unlike AFP, AREG and VEGF can that serum AREG levels were markedly increased in HCC be detected in the serum of all HCC patients. The finding that patients treated with sorafenib compared with baseline their early decrease after sorafenib treatment was correlated levels. This study was conducted on an exploratory cohort of with the efficacy of sorafenib may be of interest to clinicians. only 14 patients, and did not evaluate the prognostic role of ONCOLOGY REPORTS 41: 2041-2050, 2019 2049

AREG (16). Blivet‑Van Eggelpöel et al observed an increase Authors' contributions in serum concentrations of AREG in 10 out of 14 patients receiving sorafenib. The authors indicated that ‘no complete RC, AG and CS contributed to the study design. CG, SB, ZS, CL or partial tumour response to sorafenib was present in these and CS collected the data. CL, JCB, CG, CF, RC, AG and CS patients. Six of them showed disease progression’. Our results analyzed and interpreted the data. CS drafted the manuscript. are consistent with theirs, indicating that an increase in SB, ZS, JCB, RC and AG revised the manuscript. All authors AREG is not in favour of sorafenib efficacy. In the Blivet‑Van have read and approved the final version of the manuscript. Eggelpöel et al study, serum AREG levels were measured after 28‑458 days of sorafenib treatment. Unfortunately, the Ethics approval and consent to participate study did not provide additional dosages beyond 15 days. A complementary study would be of interest. This study was conducted in compliance with French legisla- Apart from the prognostic role of AREG and VEGF in tion and the Declaration of Helsinki regarding ethical principles HCC patients treated by sorafenib, the results of the present for medical research involving human subjects. The ‘Comité study suggest that these two biomarkers are regulated by a de Protection des Personnes d'Ile de France’ Institutional common mechanism, as a statistically significant correlation Review Board approved the study protocol (Nu‑AT140). All was observed between variations in serum AREG and VEGF patients provided their written informed consent. No samples levels. The concentrations of these two biomarkers varied in the were obtained from patients who were minors or physically or same direction (increase or decrease) in 69% of the patients after mentally unable to understand and provide their consent to the sorafenib therapy. The combination of these two biomarkers use of serum samples. was not a more discriminant predictive factor of the efficacy of sorafenib compared with each biomarker alone. This potential Patient consent for publication common pathway of regulation of AREG and VEGF may be associated with the effect of sorafenib on tumour proteostasis. Not applicable. Sorafenib alters tumour proteostasis in a complex way and via several different mechanisms (9,20,25‑27). Tumour proteostasis Competing interests includes emerging drivers of tumour progression and important determinants of clinical efficacy of cancer therapy (28). Our JCB received honoraria as a clinical consultant from Bayer. recent findings indicate that tumour markers, such as AFP, are Bayer was not involved in this study. Bayer did not provide regulated by tumour proteostasis (29,30). It is hypothesized any reagent and did not contribute to the design of this that the decrease of AREG and VEGF production may be study. All other authors declare that they have no competing associated with sorafenib‑induced alteration of proteostasis. A interests. thorough exploration of the role of proteostasis as a regulator of tumour response to sorafenib is required. References In conclusion, the present study demonstrated a broad 1. European Association for the Study of the Liver. Electronic inhibitory effect of sorafenib on the production of bioactive address: [email protected], European Association for proteins, such as AREG and VEGF, in vitro. The decrease of the Study of the Liver: EASL Clinical Practice Guidelines: serum AREG or VEGF levels was found to be associated with Management of hepatocellular carcinoma. J Hepatol 69: 182‑236, 2018. better OS and PFS in HCC patients. At this stage, the role of 2. Davis MI, Hunt JP, Herrgard S, Ciceri P, Wodicka LM, Pallares G, sorafenib in these decreases remains to be formally demon- Hocker M, Treiber DK and Zarrinkar PP: Comprehensive strated. However, this effect may be associated with the ability analysis of kinase inhibitor selectivity. Nat Biotechnol 29: 1046‑1051, 2011. of sorafenib to alter proteostasis, as shown in a previous study. 3. Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, The present study identified two biomarkers of sorafenib effi- de Oliveira AC, Santoro A, Raoul JL, Forner A, et al; SHARP cacy, detectable early in all HCC patients, and potentially of Investigators Study Group: Sorafenib in advanced hepatocellular carcinoma. N Engl J Med 359: 378‑390, 2008. interest to clinicians. 4. Cheng AL, Kang YK, Chen Z, Tsao CJ, Qin S, Kim JS, Luo R, Feng J, Ye S, Yang TS, et al: Efficacy and safety of sorafenib Acknowledgements in patients in the Asia‑Pacific region with advanced hepatocellular carcinoma: A phase III randomised, double‑blind, placebo‑controlled trial. Lancet Oncol 10: 25‑34, 2009. The authors would like to thank Momar Diouf for his help 5. Kudo M, Finn RS, Qin S, Han KH, Ikeda K, Piscaglia F, with ROC curve analyses. Baron A, Park JW, Han G, Jassem J, et al: Lenvatinib versus sorafenib in first‑line treatment of patients with unresectable hepatocellular carcinoma: A randomised phase 3 non‑inferiority Funding trial. Lancet 391: 1163‑1173, 2018. 6. Sauzay C, Petit A, Bourgeois AM, Barbare JC, Chauffert B, Galmiche A and Houessinon A: Alpha‑foetoprotein (AFP): A The authors are grateful for the financial support received multi‑purpose marker in hepatocellular carcinoma. Clin Chim from ‘Ligue Contre le Cancer’, ‘Conseil Régional des Hauts de Acta 463: 39‑44, 2016. France (Soramix)’, ‘Comité de la Somme’ and ‘CHU Amiens’. 7. Bruix J, Reig M and Sherman M: Evidence‑based diagnosis, staging, and treatment of patients with hepatocellular carcinoma. Gastroenterology 150: 835‑853, 2016. Availability of data and materials 8. Personeni N, Bozzarelli S, Pressiani T, Rimassa L, Tronconi MC, Sclafani F, Carnaghi C, Pedicini V, Giordano L and Santoro A: Usefulness of alpha‑fetoprotein response in patients treated with The datasets used and/or analysed during the present study are sorafenib for advanced hepatocellular carcinoma. J Hepatol 57: available from the corresponding author on reasonable request. 101‑107, 2012. 2050 GODIN et al: DECREASE OF AREG OR VEGF PREDICTS SORAFENIB EFFICACY IN HCC

9. Sauzay C, Louandre C, Bodeau S, Anglade F, Godin C, Saidak Z, 21. Houessinon A, François C, Sauzay C, Louandre C, Mongelard G, Fontaine JX, Usureau C, Martin N, Molinie R, et al: Protein Godin C, Bodeau S, Takahashi S, Saidak Z, Gutierrez L, et al: biosynthesis, a target of sorafenib, interferes with the unfolded Metallothionein‑1 as a biomarker of altered redox metabolism protein response (UPR) and ferroptosis in hepatocellular in hepatocellular carcinoma cells exposed to sorafenib. Mol carcinoma cells. Oncotarget 9: 8400‑8414, 2018. Cancer 15: 38, 2016. 10. Manié SN, Lebeau J and Chevet E: Cellular mechanisms of 22. Tsuchiya K, Asahina Y, Matsuda S, Muraoka M, Nakata T, endoplasmic reticulum stress signaling in health and disease. 3. Suzuki Y, Tamaki N, Yasui Y, Suzuki S, Hosokawa T, et al: Orchestrating the unfolded protein response in oncogenesis: An Changes in plasma vascular endothelial growth factor at 8 weeks update. Am J Physiol Cell Physiol 307: C901‑C907, 2014. after sorafenib administration as predictors of survival for 11. Raskopf E, Vogt A, Sauerbruch T and Schmitz V: siRNA advanced hepato‑cellular carcinoma. Cancer 120: 229‑237, 2014. targeting VEGF inhibits hepatocellular carcinoma growth and 23. Han SX, Bai E, Jin GH, He CC, Guo XJ, Wang LJ, Li M, Ying X tumor angiogenesis in vivo. J Hepatol 49: 977‑984, 2008. and Zhu Q: Expression and clinical significance of YAP, TAZ, 12. Goel HL and Mercurio AM: VEGF targets the tumour cell. Nat and AREG in hepatocellular carcinoma. J Immunol Res 2014: Rev Cancer 13: 871‑882, 2013. 261365, 2014. 13. Castillo J, Erroba E, Perugorría MJ, Santamaría M, Lee DC, 24. Zhu AX, Kang YK, Rosmorduc O, Evans TR, Santoro A, Ross P, Prieto J, Avila MA and Berasain C: Amphiregulin contributes to Gane E, Vogel A, Jeffers M, Meinhardt G, et al: Biomarker the transformed phenotype of human hepatocellular carcinoma analyses of clinical outcomes in patients with advanced hepato- cells. Cancer Res 66: 6129‑6138, 2006. cellular carcinoma treated with sorafenib with or without erlotinib 14. Xu Q, Chiao P and Sun Y: Amphiregulin in Cancer: New Insights in the SEARCH trial. Clin Cancer Res 22: 4870‑4879, 2016. for Translational Medicine. Trends Cancer 2: 111‑113, 2016. 25. Booth L, Shuch B, Albers T, Roberts JL, Tavallai M, Proniuk S, 15. Ezzoukhry Z, Louandre C, Trécherel E, Godin C, Chauffert B, Zukiwski A, Wang D, Chen CS, Bottaro D, et al: Multi‑kinase Dupont S, Diouf M, Barbare JC, Mazière JC and Galmiche A: inhibitors can associate with heat shock proteins through their EGFR activation is a potential determinant of primary resistance NH2‑termini by which they suppress chaperone function. of hepatocellular carcinoma cells to sorafenib. Int J Cancer 131: Oncotarget 7: 12975‑12996, 2016. 2961‑2969, 2012. 26. Shi YH, Ding ZB, Zhou J, Hui B, Shi GM, Ke AW, Wang XY, 16. Blivet‑Van Eggelpoël MJ, Chettouh H, Fartoux L, Aoudjehane L, Dai Z, Peng YF, Gu CY, et al: Targeting autophagy enhances Barbu V, Rey C, Priam S, Housset C, Rosmorduc O and sorafenib lethality for hepatocellular carcinoma via ER Desbois‑Mouthon C: Epidermal growth factor receptor and stress‑related apoptosis. Autophagy 7: 1159‑1172, 2011. HER‑3 restrict cell response to sorafenib in hepatocellular 27. Shimizu S, Takehara T, Hikita H, Kodama T, Tsunematsu H, carcinoma cells. J Hepatol 57: 108‑115, 2012. Miyagi T, Hosui A, Ishida H, Tatsumi T, Kanto T, et al: Inhibition 17. Livak KJ and Schmittgen TD: Analysis of relative gene expression of autophagy potentiates the antitumor effect of the multikinase data using real‑time quantitative PCR and the 2(‑Delta Delta inhibitor sorafenib in hepatocellular carcinoma. Int J Cancer 131: C(T)) Method. Methods 25: 402‑408, 2001. 548‑557, 2012. 18. Lencioni R: New data supporting modified RECIST (mRECIST) 28. Hetz C, Chevet E and Oakes SA: Erratum: Proteostasis control by for Hepatocellular Carcinoma. Clin Cancer Res 19: 1312‑1314, the unfolded protein response. Nat Cell Biol 17: 829‑838, 2015. 2013. 29. Galmiche A, Sauzay C, Houessinon A, Chauffert B and Pluquet O: 19. Abou‑Alfa GK, Schwartz L, Ricci S, Amadori D, Santoro A, Probing Tumour Proteostasis and the UPR with Serum Markers. Figer A, De Greve J, Douillard JY, Lathia C, Schwartz B, et al: Trends Cancer 2: 219‑221, 2016. Phase II study of sorafenib in patients with advanced hepato- 30. Houessinon A, Gicquel A, Bochereau F, Louandre C, Nyga R, cellular carcinoma. J Clin Oncol 24: 4293‑4300, 2006. Godin C, Degonville J, Fournier E, Saidak Z, Drullion C, et al: 20. Louandre C, Ezzoukhry Z, Godin C, Barbare JC, Mazière JC, Alpha‑fetoprotein is a biomarker of unfolded protein response Chauffert B and Galmiche A: Iron‑dependent cell death of hepa- and altered proteostasis in hepatocellular carcinoma cells tocellular carcinoma cells exposed to sorafenib. Int J Cancer 133: exposed to sorafenib. Cancer Lett 370: 242‑249, 2016. 1732‑1742, 2013.